How Secure is Tor? Not secure at all | Hacker News Reader

How Secure is Tor? Not secure at all

csam-bib.github.io

21 pointsemeryberger55 comments

  The Tor Project web site makes a bold claim to its users:

  "Tor Browser prevents someone watching your connection from knowing what websites you visit. All anyone monitoring your browsing habits can see is that you're using Tor."

  Don't misinterpret this claim. It's not true that Tor protects you against "anyone monitoring your browsing".

  On this web page, I provide you with the critical information missing from the Tor Project's website: if you estimate your adversary's resources in dollars, I'll estimated the probability that Tor will fail to protect you.

(web site by Brian Levine, Professor at UMass Amherst <https://www.cics.umass.edu/about/directory/brian-levine> and director of Rescue Lab <https://www.rescue-lab.org/>)

If an adversary is spending tens or hundreds of thousands of dollars to find you, that's a lift that most threat actors won't be able to do. Especially if they have to host a significant number of exit nodes for a lengthy period, which often means serving unlawful content which is very awkward for law enforcement.

It's definitely better than regular browsing for security, but it's not perfect.

Unfortunately, the money isn't just to find "you". You rent arbitrary exit nodes, and if you spend ~$30k / month, you'll be able to deanonimize >50% of users using Tor each month.

It’s a drop in the bucket for state actors who might want to find TORs target user base of dissidents, whistleblowers, and journalists.

It’s extremely unlikely that they would be able to find an end user (not an onion site operator, a user) with good opsec who connects occasionally, such as a journalist uploading a few documents to a secure onion drop. All existing known attacks were against onion site operators running for long periods from a static location (still took a lot of resources and time to track them down) or end users with poor opsec/infosec.

The whole thing reads as scaremongering FUD to prevent people from using Tor, with further FUD tacked on to make people think that using it might be illegal somehow. Tor is actually great for personal infrastructure (no need for domain names or a static IP), limited anonymity, and censorship resistance.

The site linked takes a shot at enumerating how unlikely it is. Do you claim it is wrong? If so, what is your calculated chance?

To me, TOR is not adequate to protect users targeted by a nation state who are the ones that TOR claims to be created for.

Given that onion sites require six hops and that the Tor team keeps watch for suspicious node behavior, and that there is no “exit node” where you can more closely observe outgoing traffic, onion connections are actually very tough to correlate. It requires a large number of compromised nodes plus cooperation with ISPs and backbone providers, as seen in the arrest of the onion site operator a few years ago. There were some good writeups at the time. You basically need to use DDoS techniques combined with targeted disconnections to narrow down the list of potential targets, even while owning many nodes. And onions have seen some DDoS hardening since this time.

Clearnet traffic via exit node is a bit different. With only three hops it might be possible to correlate targeted traffic by owning a huge number of nodes, but even then, unless you also control the server being connected (or it barely receives any traffic) then it may not give you anything actionable. (Using Tor is not a crime.) Unless you can see what is being done on the server by the unmasked user, or you can establish a pattern of behavior, or you see something like a large data transfer whose size matches a known event of interest, then all you know is someone accessed the server over Tor. And even then, owning both the entry and exit isn’t sufficient if the user is masking their traffic with decoy and/or relay traffic.

But you don't need all the hops. You can run a correlation attack (which has been long known):

https://github.com/Attacks-on-Tor/Attacks-on-Tor

and if you can get the guard and exit node for a clearnet connection and the guard, rendezvous point and exit for the onion service that can be enough.

You ignored a substantial portion of the reply. “That can be enough”... yet it has not been. It’s actually very difficult to perform correlation attacks in a complex network, especially if the user is generating decoy traffic or passing along relay traffic, and even moreso if the end server is highly active as well. It takes an enormous amount of resources to even determine that someone may be connecting to a specific server, and again just connecting to a specific server often tells you nothing about what the user was specifically doing. So unless you’re hunting down someone selling enriched uranium, major abuse content producers/hosters, or something of that scale, putting in all that investment to gain a fuzzy data point that likely isn’t even useful in an enforcement context is just not worth the tradeoff.

Come back when you have evidence of real-world attacks and not just FUD against the best current network for anonymity.

> So unless you’re hunting down someone selling enriched uranium, major abuse content producers/hosters, or something of that scale, putting in all that investment to gain a fuzzy data point that likely isn’t even useful in an enforcement context is just not worth the tradeoff.

But I don’t think we disagree. My view is that TOR is inadequate against a nation state attack because for some of these attacks it is easier to do mass de-anonymization and hope you get some particular user or set of users you are interested in. The resources to do this are small for something the scale of an intelligence agency, but excessively large for some local police department.

I’m not sure why you appear so hostile to citing attacks that are well-known and already part of the public threat model.

Even mass deanonymization—which is unlikely to be sustainable long-term in such an adversarial ecosystem—does nothing for you unless you (a) know specifically what you are looking for and (b) are able to either compromise the end server or identify frequent or very large transfers that correspond exactly with known events or data of interest (like specific uploads). There are some countries where the authorities might be allowed to round up anyone who has connected to a server without further information about what they were doing, but despite our declining civil liberties situation, these countries aren’t currently in the West.

There just aren’t that many people who are both legitimate and likely targets of such an attack. And since the most likely actor to be able to afford such an attack (USG) also has practical uses for Tor, IMHO it would be unlikely to do anything that actually threatens the network. I could be misremembering, but I believe the one big successful deanonymization attack was in Europe, not the US, and the approach used there would not have worked to locate an occasional end user of a busy server.

I am not really interested in debating this further. Feel free to respond of course, but it’s obvious to me (and hopefully everyone else) that you have an axe to grind against Tor.

I'm not here to defend Tor

But the calculator states that if the investigating party has $150,000 a month budget for all targets they have a 100% certainty of getting your IP address... obviously this is false, so what else has the author claimed that is also not true?

Pretty much everything claimed on this site is false or grossly misleading.

Really? Tell me why.

The primary claims of the site, both made without any evidence (presumably by you), are that

1. Tor is primarily used to distribute CSAM,

2. a single organization with a budget of $150k could deanonymize every Tor user simultaneously.

Since pretty much every firat world law enforcement organization can cough up this amount in spare budget, either

- at least one of the claims above is false; or

- there's a global conspiracy involving every major law enforcement organization in the planet being taken over by pedos.

In fact, both claims (you?) made without evidence are simply false.

Having published calculations for the second claim is like having published calculations for "the Sun went supernova yesterday". The conclusion is blatantly wrong, so the calculations have a mistake, and an intellectually honest author would double check them, find that mistake, then retract the claim (or would not have made it in the first place).

1. I said “extensively” used for csam. What’s my source/evidence for that claim? This list of peer reviewed papers, cases, and government reports: https://csam-bib.github.io.

2. My site shows a mathematical model of security that Tor provides in terms of its design for relays alone. I say on the site I’m not including staff and other costs. In fact bringing someone to court is a further cost. My point in making the site is to quantify solely the costs that the design brings to the table. You can then compare that design to some other anonymous system. Or compare it to a doublespend attack on bitcoin or to brute force decryption. That’s important for users.

Unlike the Tor Project, I’m being transparent by showing assumptions, the math, and the code. Do you have a better model? Great, then publish it. I’m trying to start a formal conversation. The Tor Project should be relying on science, and not strong assertions, to ensure its security.

And while there are costs to, say, bring someone to court for csam, do you believe all adversaries are going to do that? That’s why it’s not part of the costs I model.

Finally, to be more clear, Onion Services in particular are the problem when it comes to CSAM (and ransomeware). Tor Browser is not the issue when it comes to CSAM.

Not only is it misleading, but given how it’s presented, it’s clearly FUD in the interest of the author’s pet cause (campaigning against Tor use due to a perceived association with CSAM).

Tor isn't without its weaknesses, but this author is simultaneously claiming child predators are successfully evading law enforcement despite their identity only coming at the relatively low price of ~$150k.

The math and the code is all there. I’d love to have a discussion about what the real value is. Further, why hasn’t the Tor Project provided this calculation? Why hasn’t anyone? I think it’s necessary.

The assumption is the adversary controls x of N nodes. When x=N the probability of discovering the onion service IP is 1. But the adversary can not achieve this situation as he only controls the additional nodes. The existing nodes still stay in the network, they do not disappear. The ratio is not x/N but x/(x+N).

The formula is wrong and it all falls apart.

You can adjust the code on the page easily (it’s open source javascript) to determine the question you are after, which is a valid one: if an adversary starts today and adds x nodes to the existing network, what is their success rate?

BUT the author asked a different (but valid) question: assuming the adversary controls x out of N existing nodes, what is the success rate? I am unclear: is the assertion that everyone’s relay is honest today? From a privacy standpoint, that’s not a great assumption.

No, the author is presenting an idea that $25 a month can buy you a node. That fits adding a new node to the network, not taking over an existing node.

I am the author. I am telling you are wrong about that.

We are all, in unison, saying you -the author- is the one who is wrong.

Posting some words on a URL does not make them factually accurate.

"As C3P will tell you: CSAM distribution on Tor onion services is not inevitable."

Lol, are we using the regular internet as an example of preventing all CSAM?

We've known for years that owning enough nodes results in the compromise of privacy and that it's likely the NSA has achieved this. Although there is some question around how that plays out if adversaries like China are also competing for similar node share percentage.

There is no question about that. The site makes use of current statistics from the Tor Project.

As a percent of onion services, what does it work out to, a few percent? And how much of that is dedicated abuse sites versus general adult sites?

Your question made me curious so I tried to see what information about onion sites is available. It’s hard to measure onions sites by design, but

https://99firms.com/research/tor-stats

Says there seem to be about 65k onion sites.

This site:

https://protectchildren.ca/en/press-and-media/blog/2025/tor-...

Has some varying numbers depending on the observation time, but in final month listed saw 30k sites that had they identified as having CSAM.

I’m not sure how accurate either number is or if they are directly comparable but that would be a 50% of all onion sites ballpark.

Not sure how to measure general sites vs dedicated abuse sites.

Looks like the tor metrics site says ~900k onion sites?

https://metrics.torproject.org/hidserv-dir-v3-onions-seen.ht...

In reality there are fewer than 1000 "real" hidden services (see https://rnsaffn.com/zg4/ for HTTP response dumps) and an ocean of short-lived temporary spam onions, most of them pornography.

Correct me if I'm wrong, but this feels like a long-winded way of saying: if an adversary could control a significant portion of relays without being found out and for a not-insignificant period of time, it could defeat Tor.

Is it correct? Probably. Does it justify the "Not secure at all" indictment? No.

The calculator also misleads in another direction, in that it could underestimate the probability of failure by only considering the "takeover" scenario, while I think it is much more likely to be defeated via other OpSec failures.

The “conclusions” section of the site makes this same point.

The website actually states “not very secure at all”. This hacker news submission changed the title.

I wouldn't use Tor or any other anonymous services like SecureDrop without a VPN (preferably multi-hop). Otherwise you're advertising to the world that your IP address uses Tor, and that alone can be a huge reduction in the solution space for your adversary to deanoymize you.

I agree, but we are both first-world privileged.

How exactly does someone in China or North Korea go about getting a multi-hop VPN to access Tor?

"The small set of people that centrally control Tor software and centrally manage the Tor network have the power to act to stop this abuse without lessening their (weak) protections."

That the author has received funding from the DOJ makes me wonder what their proposed solution is.

> Why has the Tor Project created a network used extensively for child sexual abuse

Hottest take of the week right there.

Why do they seem to imply that Tor was somehow created explicitly with this purpose in mind? That's like saying only criminals use the Internet, just because it can be used to commit crimes.

I think they are taking Tor's words and applying it to a much broader scope than they originally intended.

> Tor Browser prevents someone watching your connection from knowing what websites you visit.

If someone is watching only your connection as it exits your local ISP and nothing else, then yes, this is in fact true. It's just not articulated that plainly.

But if the author actually went as far as they are trying to, they might as well tell people to just give up because there's a chance your attacker already controls the destination server you're talking to in the first place.

If you're going to the trouble of trying to calculate the chances that nodes in the middle are compromised, why not include the destination itself too?

> The small set of people that centrally control Tor software and centrally manage the Tor network have the power to act to stop this abuse without lessening their (weak) protections.

Source: trust me bro

> The world's standards for encrypting data are so secure that no one has enough money or time to brute force their way into properly encrypted data, not even governments. They are better off waiting for a scientific breakthrough that may never come.

This completely disregards the possibility that any one of a number of root CAs aren't already compromised or cannot be coerced by your attacker.

If you're going to claim tor is insecure, you might as well go all the way and say it's pointless to use anything at all, ever.

> Why has the US created a highway system used extensively for causing death?

Reminds me of https://0x0.st/XJZT.jpg

My apologies. I don’t believe that was their intent to create a network for csam. But after decades of it being used extensively for csam, why would they take no corrective action?

Maybe because there isn't a known solution?

CSAM is still distributed on the clearnet too... why isn't there a "solution" for that too?

So far the only solutions people seem to have come up with is mass surveillance, and that's not an option.

There is a known solution.

Did you know that the Tor Project allows exit nodes to filter based on the clear internet IP. So filtering is ok.

However, if a relay refuses to service an onion site directory look up, it will be banned by the Directory Authority. They could allow this today. But they don’t. That’s the simple solution. No surveillance. Not back door. No less privacy for everyone else.

edit: This is easy to confirm. I’m not asking anyone to trust me.

Exit nodes are not used for onion services. From https://onionservices.torproject.org/technology/properties/:

> For the Tor network, Onion Services can alleviate the load on exit nodes, since it's connections don't need to reach the exits.

Also:

> Directory Authority.

"These authorities are operated by trusted organizations or individuals with a strong commitment to the principles of privacy, security, and network neutrality."

Emphasis on neutrality... it's not the job of network operators to police the sites people can and can't access, this is exactly why many people use Tor in the first place.

> They could allow this today. But they don’t.

Speaking for onion services... no, they cannot, because the entire design of the tor network prevents this in the first place. No relay in the circuit knows the final destination because it is encrypted multiple times (like an onion) and each hop can only see where it needs to go next, not what the destination is.

I think the point is that exit node operators can filter traffic they don’t want to support. Guard and middle nodes are not given the same choice; they apparently must support all traffic or get booted. Why can’t other nodes have freedom to decide how they want to participate?

> Why can’t other nodes have freedom to decide how they want to participate?

Because the network was explicitly designed to not allow this... otherwise it becomes subject to censorship, which is one of the main goals they try to prevent.

The (onion) address itself is never transmitted in plaintext through the Tor network... when you access an onion site, your Tor client encrypts the traffic multiple times, literally like an onion. No relay in the circuit knows the final destination.

It is absolutely a design decision. I don’t understand though how allowing exit nodes to filter (by port and IP) doesn’t permit censorship but allowing internal nodes to not complete connections to onion sites does. I do understand that early nodes on the path are unaware of what the traffic but it seems pretty straightforward to allow nodes to not become rendezvous points for onion sites.

You are welcome to fork Tor and create a version that uses this approach, but good luck getting people to use it.

Conversely, even if the official project implemented an onion blacklist, a fork would quickly appear to remove it. And node operators would likely prefer that one.

Anyone with any sense understands that introducing a node blacklist creates the capability to expand the use of that blacklist in the interest of political and/or military censorship. The Tor project, Tor devs, and node operators are adamantly opposed to any such censorship capabilities. Therefore it will not happen, period.

That’s not at all what I proposed. Not even close.

Edit: on second look I can see how you could think it was. I’m just proposing that if you run a node you be allowed to not become a rendezvous point for onion sites.

Ah, I misunderstood. In any case, onion traffic is not flagged as such, so nobody but the last hop would even know that onion traffic is being passed. Allowing the last node to kill the whole circuit seems like it would cause routing problems and/or contribute towards deanonymization, as would flagging onion traffic at every step of the journey.

Your assumptions are based on faulty understanding of how tor works.

I understand well how it works. I agree this is not possible today’s code base but that limitation is due to a design choice. It’s due to a policy decision that the privacy of children who have been sexually exploited is not as important as the privacy of others (including the privacy of people who sexually exploit children). It’s not a technical limitation. It’s a flaw.

Specifically, it would be easy to add code to hsdir functionality to deny requests for onion sites that are known to be related to csam. Those sites could be announced by the DAs as part of the consensus file, for example. The Tor Project currently lets exit nodes filter by IP address as long as they announce that in their config; this new functionality is of the same kind in the abstract. This change would not be a backdoor. It’s not going to weaken the privacy of anyone using Tor.

The current setup is an extremist position that children who have been abused are not deserving of privacy. It’s a position that all information deserves to be free even if that information is very clearly harmful to others and has no positive benefit to society. One can have that opinion but you won’t find many (outside of this thread) that agree.

>Did you know that the Tor Project allows exit nodes to filter based on the clear internet IP. So filtering is ok.

That's simply not true. Exit operators who intentionally block websites are flagged as bad relays.

https://community.torproject.org/policies/relays/expectation... https://gitlab.torproject.org/tpo/network-health/team/-/wiki...

I think that is referring to the node exit policy, which explicitly allows particular ports and IP addresses to be blocked.

https://support.torproject.org/relay-operators/exit-policies...

The documents you referred to just say you need to honor your own exit policy.

Clickbait title is usually a good indicator of clickbait content.

I see in the comments that the author is an academic, my cursory look of the site makes me disappointed to see such weak rigor applied here. This looks like a hit piece dressed up to sound scary. Not going to waste my time further on its claims when on the surface its given me this impression. Strikes me as yelling and not listening type of personality.

Reply on news.ycombinator.com